There are at least two separate networks are found in Data Centers. The more ubiquitous of the networks, the Local Area Network (LAN) based on the Ethernet protocol, is mainly used for server to server and server to Internet communications. The other network, the Storage Area Network (SAN), is specialized to carry server to storage communications. The Data Center Storage Area Network is mainly based on the Fibre Channel protocol and has the following characteristics: low latency, high bandwidth, and a loss-less network. Recently there have been innovations to merge the Storage Area Network (SAN) with the Local Area Network. The promised benefits include a savings on the amount of equipment and the resulting savings on the amount of equipment real estate, power, and cooling required. Newly created standards comprising this LAN/SAN convergence define how Storage Area Network frames, namely Fibre Channel protocol frames, are mapped over the Ethernet network. These new frames are called Fibre Channel over Ethernet (FCoE) network frames. Additional standards define how to make the Ethernet network lossless, i.e., to add flow control at the network level to prevent Ethernet frames from being dropped due to congestion. Still other standards define how to segment the transmission line into classes that virtually separate the communications over the transmission line.
Converging the LAN and SAN networks has created additional complexity in the management, control, and data switching areas. Singly, the Fibre Channel switch fabric protocols are very complex and have shown to be not very interoperable between the small number of vendors who build products that support them. Mapping the Fibre Channel switch fabric protocols over Ethernet has resulted in a dizzying amount of new standards that have inhibited the market acceptance of the Fibre Channel over Ethernet (FCoE) mapping over this new converged network. New switches have been defined called Fibre Channel Forwarders (FCFs) and Fibre Channel Data Forwarders (FDFs), which add Fibre Channel over Ethernet and Ethernet elements to the already complex Fibre Channel switch architecture. FCFs and FDFs interconnect ENodes, which are Fibre Channel or devices nodes that are able to transmit Fibre Channel over Ethernet frames. There have been some standards and innovations applied to ENodes, and their embedded Virtual N_Ports (VN_Ports), to connect without using FCF's or FDF's. One of these efforts defines an ENode to ENode connection method, called VN_Port to VN_Port (VN2VN) whereby ENodes can connect to each other over a Lossless Ethernet network without an FCF or FDF. Other methods have been suggested to move some of the FCF/FDF intelligence to the ENode. Both the emerging VN2VN standard and the emerging direct ENode direct connect methods have many significant disadvantages. These disadvantages include but are not limited to: the requirement for the ENode to choose a unique MAC address for each VN_Port, the requirement for the ENode to choose a unique Fibre Channel address identifier for each VN_Port, the lack of visibility into the network's supported maximum frame size or other capabilities, the lack of standardized discovery of specific ENode types such as Storage targets, the lack of the ability to automatically and dynamically create Fibre Channel zones or access control lists (ACLs) for intermediate Ethernet bridges, the lack of visibility to load balance across several paths from a source ENode to a destination ENode based on FCID's, and the increased complexity to scale to hundreds of ENodes which requires error prone manual configuration. Due to the lack of Fibre Channel fabric control, these emerging ideas and standards target smaller networks of ENodes, which are impractical in today's Data Center.
In parallel with the innovations around converging the LAN and SAN, there have also been a trend to virtualize servers, i.e., consolidate a corporation's many underutilized servers onto fewer more utilized servers. The server visualization trend has many advantages, including more utilization of existing underutilized servers, lower equipment space, power, and cooling requirements since there are fewer servers. This trend results in fewer and higher utilized servers which have changed the traffic characteristics of the Local Area Network that interconnects them. The traffic requirements which used to be flowing from Internet to Server have changed to an any-to-any server flow. This migration in traffic patterns has produced a trend to “flatten”LANs, i.e., consolidate the normally three layers (core, distribution, and access) of switches commonly found in a Data Center to two layers (core and access). In parallel with this physical flattening trend is the trend towards utilizing layer 2 forwarding methods to keep the network in a single broadcast domain, which helps support any-to-any connection requirements of virtualized servers and their hypervisors. New link level protocols have been defined to accelerate the ability for any to any server based virtual machine communications. Many of these new link level protocols need new switch hardware and new ways to manage the resulting network.
What is needed is a simpler way to converge the LAN and SAN in a scalable and less complex method than the trajectory of both the standards committees and emerging ENode to ENode inventions. What is also needed is have this simpler method be more compatible with the trend towards flattening the large Data Center networks. Both simpler methods need to be easily managed, scalable, and interoperable. Accomplishing this would accelerate LAN/SAN network convergence trend and accelerate the flattening of the LAN to more easily attain the benefits of visualization, convergence, and consolidation.